The Experts below are selected from a list of 46563 Experts worldwide ranked by ideXlab platform
Jie Zhou - One of the best experts on this subject based on the ideXlab platform.
-
Learning Discriminative Aggregation Network for Video-Based Face Recognition and Person Re-identification
International Journal of Computer Vision, 2018Co-Authors: Yongming Rao, Jie ZhouAbstract:In this paper, we propose a discriminative Aggregation Network method for video-based face recognition and person re-identification, which aims to integrate information from video frames for feature representation effectively and efficiently. Unlike existing video Aggregation methods, our method aggregates raw video frames directly instead of the features obtained by complex processing. By combining the idea of metric learning and adversarial learning, we learn an Aggregation Network to generate more discriminative images compared to the raw input frames. Our framework reduces the number of image frames per video to be processed and significantly speeds up the recognition procedure. Furthermore, low-quality frames containing misleading information can be well filtered and denoised during the Aggregation procedure, which makes our method more robust and discriminative. Experimental results on several widely used datasets show that our method can generate discriminative images from video clips and improve the overall recognition performance in both the speed and the accuracy for video-based face recognition and person re-identification.
-
ICCV - Learning Discriminative Aggregation Network for Video-Based Face Recognition
2017 IEEE International Conference on Computer Vision (ICCV), 2017Co-Authors: Yongming Rao, Ji Lin, Jie ZhouAbstract:In this paper, we propose a discriminative Aggregation Network (DAN) method for video face recognition, which aims to integrate information from video frames effectively and efficiently. Unlike existing Aggregation methods, our method aggregates raw video frames directly instead of the features obtained by complex processing. By combining the idea of metric learning and adversarial learning, we learn an Aggregation Network that produces more discriminative synthesized images compared to raw input frames. Our framework reduces the number of frames to be processed and significantly speed up the recognition procedure. Furthermore, low-quality frames containing misleading information are filtered and denoised during the Aggregation process, which makes our system more robust and discriminative. Experimental results show that our method can generate discriminative images from video clips and improve the overall recognition performance in both the speed and accuracy on three widely used datasets.
Gang Hua - One of the best experts on this subject based on the ideXlab platform.
-
gated context Aggregation Network for image dehazing and deraining
Workshop on Applications of Computer Vision, 2019Co-Authors: Dongdong Chen, Qingnan Fan, Jing Liao, Liheng Zhang, Dongdong Hou, Lu Yuan, Gang HuaAbstract:Image dehazing aims to recover the uncorrupted content from a hazy image. Instead of leveraging traditional low-level or handcrafted image priors as the restoration constraints, e.g., dark channels and increased contrast, we propose an end-to-end gated context Aggregation Network to directly restore the final haze-free image. In this Network, we adopt the latest smoothed dilation technique to help remove the gridding artifacts caused by the widely-used dilated convolution with negligible extra parameters, and leverage a gated sub-Network to fuse the features from different levels. Extensive experiments demonstrate that our method can surpass previous state-of-the-art methods by a large margin both quantitatively and qualitatively. In addition, to demonstrate the generality of the proposed method, we further apply it to the image deraining task, which also achieves the state-of-the-art performance.
-
WACV - Gated Context Aggregation Network for Image Dehazing and Deraining
2019 IEEE Winter Conference on Applications of Computer Vision (WACV), 2019Co-Authors: Dongdong Chen, Qingnan Fan, Jing Liao, Liheng Zhang, Dongdong Hou, Lu Yuan, Gang HuaAbstract:Image dehazing aims to recover the uncorrupted content from a hazy image. Instead of leveraging traditional low-level or handcrafted image priors as the restoration constraints, e.g., dark channels and increased contrast, we propose an end-to-end gated context Aggregation Network to directly restore the final haze-free image. In this Network, we adopt the latest smoothed dilation technique to help remove the gridding artifacts caused by the widely-used dilated convolution with negligible extra parameters, and leverage a gated sub-Network to fuse the features from different levels. Extensive experiments demonstrate that our method can surpass previous state-of-the-art methods by a large margin both quantitatively and qualitatively. In addition, to demonstrate the generality of the proposed method, we further apply it to the image deraining task, which also achieves the state-of-the-art performance.
-
Gated Context Aggregation Network for Image Dehazing and Deraining
arXiv: Computer Vision and Pattern Recognition, 2018Co-Authors: Dongdong Chen, Qingnan Fan, Jing Liao, Liheng Zhang, Dongdong Hou, Lu Yuan, Gang HuaAbstract:Image dehazing aims to recover the uncorrupted content from a hazy image. Instead of leveraging traditional low-level or handcrafted image priors as the restoration constraints, e.g., dark channels and increased contrast, we propose an end-to-end gated context Aggregation Network to directly restore the final haze-free image. In this Network, we adopt the latest smoothed dilation technique to help remove the gridding artifacts caused by the widely-used dilated convolution with negligible extra parameters, and leverage a gated sub-Network to fuse the features from different levels. Extensive experiments demonstrate that our method can surpass previous state-of-the-art methods by a large margin both quantitatively and qualitatively. In addition, to demonstrate the generality of the proposed method, we further apply it to the image deraining task, which also achieves the state-of-the-art performance. Code has been made available at this https URL.
-
CVPR - Neural Aggregation Network for Video Face Recognition
2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2017Co-Authors: Jiaolong Yang, Peiran Ren, Dongqing Zhang, Dong Chen, Fang Wen, Gang HuaAbstract:This paper presents a Neural Aggregation Network (NAN) for video face recognition. The Network takes a face video or face image set of a person with a variable number of face images as its input, and produces a compact, fixed-dimension feature representation for recognition. The whole Network is composed of two modules. The feature embedding module is a deep Convolutional Neural Network (CNN) which maps each face image to a feature vector. The Aggregation module consists of two attention blocks which adaptively aggregate the feature vectors to form a single feature inside the convex hull spanned by them. Due to the attention mechanism, the Aggregation is invariant to the image order. Our NAN is trained with a standard classification or verification loss without any extra supervision signal, and we found that it automatically learns to advocate high-quality face images while repelling low-quality ones such as blurred, occluded and improperly exposed faces. The experiments on IJB-A, YouTube Face, Celebrity-1000 video face recognition benchmarks show that it consistently outperforms naive Aggregation methods and achieves the state-of-the-art accuracy.
-
Neural Aggregation Network for Video Face Recognition
arXiv: Computer Vision and Pattern Recognition, 2016Co-Authors: Jiaolong Yang, Peiran Ren, Dongqing Zhang, Dong Chen, Fang Wen, Gang HuaAbstract:This paper presents a Neural Aggregation Network (NAN) for video face recognition. The Network takes a face video or face image set of a person with a variable number of face images as its input, and produces a compact, fixed-dimension feature representation for recognition. The whole Network is composed of two modules. The feature embedding module is a deep Convolutional Neural Network (CNN) which maps each face image to a feature vector. The Aggregation module consists of two attention blocks which adaptively aggregate the feature vectors to form a single feature inside the convex hull spanned by them. Due to the attention mechanism, the Aggregation is invariant to the image order. Our NAN is trained with a standard classification or verification loss without any extra supervision signal, and we found that it automatically learns to advocate high-quality face images while repelling low-quality ones such as blurred, occluded and improperly exposed faces. The experiments on IJB-A, YouTube Face, Celebrity-1000 video face recognition benchmarks show that it consistently outperforms naive Aggregation methods and achieves the state-of-the-art accuracy.
Yongming Rao - One of the best experts on this subject based on the ideXlab platform.
-
Learning Discriminative Aggregation Network for Video-Based Face Recognition and Person Re-identification
International Journal of Computer Vision, 2018Co-Authors: Yongming Rao, Jie ZhouAbstract:In this paper, we propose a discriminative Aggregation Network method for video-based face recognition and person re-identification, which aims to integrate information from video frames for feature representation effectively and efficiently. Unlike existing video Aggregation methods, our method aggregates raw video frames directly instead of the features obtained by complex processing. By combining the idea of metric learning and adversarial learning, we learn an Aggregation Network to generate more discriminative images compared to the raw input frames. Our framework reduces the number of image frames per video to be processed and significantly speeds up the recognition procedure. Furthermore, low-quality frames containing misleading information can be well filtered and denoised during the Aggregation procedure, which makes our method more robust and discriminative. Experimental results on several widely used datasets show that our method can generate discriminative images from video clips and improve the overall recognition performance in both the speed and the accuracy for video-based face recognition and person re-identification.
-
ICCV - Learning Discriminative Aggregation Network for Video-Based Face Recognition
2017 IEEE International Conference on Computer Vision (ICCV), 2017Co-Authors: Yongming Rao, Ji Lin, Jie ZhouAbstract:In this paper, we propose a discriminative Aggregation Network (DAN) method for video face recognition, which aims to integrate information from video frames effectively and efficiently. Unlike existing Aggregation methods, our method aggregates raw video frames directly instead of the features obtained by complex processing. By combining the idea of metric learning and adversarial learning, we learn an Aggregation Network that produces more discriminative synthesized images compared to raw input frames. Our framework reduces the number of frames to be processed and significantly speed up the recognition procedure. Furthermore, low-quality frames containing misleading information are filtered and denoised during the Aggregation process, which makes our system more robust and discriminative. Experimental results show that our method can generate discriminative images from video clips and improve the overall recognition performance in both the speed and accuracy on three widely used datasets.
James Z. Wang - One of the best experts on this subject based on the ideXlab platform.
-
Deep multi-patch Aggregation Network for image style, aesthetics, and quality estimation
Proceedings of the IEEE International Conference on Computer Vision, 2015Co-Authors: Xin Lu, Radomir Mech, Xiaohui Shen, Zhe Lin, James Z. WangAbstract:This paper investigates problems of image style, aes-thetics, and quality estimation, which require fine-grained details from high-resolution images, utilizing deep neural Network training approach. Existing deep convolutional neural Networks mostly extracted one patch such as a downsized crop from each image as a training example. However, one patch may not always well represent the entire image, which may cause ambiguity during training. We propose a deep multi-patch Aggregation Network training approach, which allows us to train models using multiple patches generated from one image. We achieve this by constructing multiple, shared columns in the neural Network and feeding multiple patches to each of the columns. More importantly, we propose two novel Network layers (statistics and sorting) to support Aggregation of those patches. The proposed deep multi-patch Aggregation Network integrates shared feature learning and Aggregation function learning into a unified framework. We demonstrate the effectiveness of the deep multi-patch Aggregation Network on the three problems, i.e., image style recognition, aesthetic quality categorization, and image quality estimation. Our models trained using the proposed Networks significantly outperformed the state of the art in all three applications.
-
ICCV - Deep Multi-patch Aggregation Network for Image Style, Aesthetics, and Quality Estimation
2015 IEEE International Conference on Computer Vision (ICCV), 2015Co-Authors: Zhe Lin, Radomir Mech, Xiaohui Shen, James Z. WangAbstract:This paper investigates problems of image style, aesthetics, and quality estimation, which require fine-grained details from high-resolution images, utilizing deep neural Network training approach. Existing deep convolutional neural Networks mostly extracted one patch such as a down-sized crop from each image as a training example. However, one patch may not always well represent the entire image, which may cause ambiguity during training. We propose a deep multi-patch Aggregation Network training approach, which allows us to train models using multiple patches generated from one image. We achieve this by constructing multiple, shared columns in the neural Network and feeding multiple patches to each of the columns. More importantly, we propose two novel Network layers (statistics and sorting) to support Aggregation of those patches. The proposed deep multi-patch Aggregation Network integrates shared feature learning and Aggregation function learning into a unified framework. We demonstrate the effectiveness of the deep multi-patch Aggregation Network on the three problems, i.e., image style recognition, aesthetic quality categorization, and image quality estimation. Our models trained using the proposed Networks significantly outperformed the state of the art in all three applications.
Achille Pattavina - One of the best experts on this subject based on the ideXlab platform.
-
Optimal BBU placement for 5G C-RAN deployment over WDM Aggregation Networks
Lightwave Technology, Journal of, 2016Co-Authors: Francesco Musumeci, Camilla Bellanzon, Nicola Carapellese, Massimo Tornatore, Achille Pattavina, Stephane GosselinAbstract:5G mobile access targets unprecedented performance, not only in terms of higher data rates per user and lower latency, but also in terms of Network intelligence and capillarity. To achieve this, 5G Networks will resort to solutions as small cell deployment, multipoint coordination (CoMP, ICIC) and Centralized Radio Access Network (C-RAN) with Baseband Units (BBUs) hotelling. As adopting such techniques requires a high-capacity low-latency access/Aggregation Network to support backhaul, radio coordination and fronthaul (i.e., digitized baseband signal) traffic, optical access/Aggregation Networks based onWavelength Division Multiplexing (WDM) are considered as an outstanding candidate for 5G-transport. By physically separating BBUs from the corresponding cell sites, BBU Hotelling promises substantial savings in terms of cost and power consumption. However, this requires to insert additional high bit-rate traffic, i.e., the fronthaul, which also has very strict latency requirements. Therefore, a trade-off between the number of BBUhotels (BBU consolidation), the fronthaul latency and Network-capacity utilization arises. We introduce the novel BBU-placement optimization problem for CRAN deployment over a WDM Aggregation Network and formalize it by Integer Linear Programming. Thus, we evaluate the impact of i) jointly supporting converged fixed and mobile traffic, ii) different fronthaultransport options (namely, OTN and Overlay) and iii) joint optimization of BBU and electronic switches placement, on the amount of BBU consolidation achievable on the Aggregation Network.
-
BBU placement over a WDM Aggregation Network considering OTN and overlay fronthaul transport
2015 European Conference on Optical Communication (ECOC), 2015Co-Authors: Nicola Carapellese, Massimo Tornatore, Achille Pattavina, Stephane GosselinAbstract:We formalize a BBU placement problem for C-RAN infrastructures based on WDM Aggregation Networks, considering both OTN and Overlay for fronthaul transport. Numerical simulations obtained over realistic instances show that Overlay enables higher BBU consolidation (fewer BBU hotel sites).
-
On the Placement of BBU Hotels in an Optical Access/Aggregation Network for 5G Transport
Asia Communications and Photonics Conference 2015, 2015Co-Authors: Francesco Musumeci, Camilla Bellanzon, Nicola Carapellese, Massimo Tornatore, Achille Pattavina, Stephane GosselinAbstract:We discuss the placement of BBU-Hotels over optical access/Aggregation Networks for 5G backhaul, comparing OTN and Overlay fronthaul solutions. Different capacity and latency constraints lead to non-univocal optimal choices to minimize the number of BBU-Hotels.
-
energy efficient baseband unit placement in a fixed mobile converged wdm Aggregation Network
IEEE Journal on Selected Areas in Communications, 2014Co-Authors: Nicola Carapellese, Massimo Tornatore, Achille PattavinaAbstract:Energy efficiency is expected to be a key design parameter for next-generation access/Aggregation Networks. Using a single Network infrastructure to aggregate/backhaul both mobile and fixed Network traffic, typically referred to as fixed/mobile convergence (FMC), seems a promising strategy to pursue energy efficiency. WDM Networks are a prominent candidate to support next-generation FMC Network architectures, as they provide huge capacity at relatively low costs and energy consumption. We consider a FMC WDM Aggregation Network in which the novel concept of “hotelling” of mobile baseband units (BBUs) is employed. The so-called BBU hotelling consists in separating BBUs from their cell sites and consolidating them in single locations, called hotels. As a result, the Aggregation Network will transport both IP (fixed and mobile) traffic and CPRI (fronthaul) traffic, the latter exchanged between each BBU and its cell site. In this paper, we propose an energy-efficient WDM Aggregation Network, and we formally define the BBU placement optimization problem, whose objective is to minimize the defined Aggregation infrastructure power (AIP). We consider three different Network architectures: Bypass, Opaque, and No-Hotel, which feature different placement of BBUs and routing of traffic. By modeling the power contributions of each active device, we study how and how much BBU consolidation, optical bypass, and active traffic Aggregation influence the AIP, for all the three architectures. Our results show that, in our case study, the proposed architectures enable savings up to about 60%-65% in dense-urban/urban and about 40% in rural scenarios.
-
energy efficient baseband unit placement in a fixed mobile converged wdm Aggregation Network
IEEE Journal on Selected Areas in Communications, 2014Co-Authors: Nicola Carapellese, Massimo Tornatore, Achille PattavinaAbstract:Energy efficiency is expected to be a key design parameter for next-generation access/Aggregation Networks. Using a single Network infrastructure to aggregate/backhaul both mobile and fixed Network traffic, typically referred to as fixed/mobile convergence (FMC), seems a promising strategy to pursue energy efficiency. WDM Networks are a prominent candidate to support next-generation FMC Network architectures, as they provide huge capacity at relatively low costs and energy consumption. We consider a FMC WDM Aggregation Network in which the novel concept of “hotelling” of mobile baseband units (BBUs) is employed. The so-called BBU hotelling consists in separating BBUs from their cell sites and consolidating them in single locations, called hotels. As a result, the Aggregation Network will transport both IP (fixed and mobile) traffic and CPRI (fronthaul) traffic, the latter exchanged between each BBU and its cell site. In this paper, we propose an energy-efficient WDM Aggregation Network, and we formally define the BBU placement optimization problem, whose objective is to minimize the defined Aggregation infrastructure power (AIP). We consider three different Network architectures: Bypass, Opaque, and No-Hotel, which feature different placement of BBUs and routing of traffic. By modeling the power contributions of each active device, we study how and how much BBU consolidation, optical bypass, and active traffic Aggregation influence the AIP, for all the three architectures. Our results show that, in our case study, the proposed architectures enable savings up to about 60%-65% in dense-urban/urban and about 40% in rural scenarios.
